Update src/qa.py

src/qa.py

@@ -125,16 +125,30 @@ def generate_answer(query: str, retrieved_chunks: list):
     prompt = PROMPT_TEMPLATE.format(context=context, query=query)
 
     try:
-        result = _answer_model(
-            prompt,
-            max_new_tokens=300,
-            do_sample=False,
-            temperature=0.2
+    result = _answer_model(
+        prompt,
+        max_new_tokens=350,        # allow longer, more complete answers
+        do_sample=True,            # enable sampling for natural flow
+        temperature=0.7,           # slightly higher = more expressive responses
+        top_p=0.95,                # nucleus sampling for coherence
+        repetition_penalty=1.2     # discourages repetitive phrasing
+    )
+
+    answer = result[0]["generated_text"].strip()
+
+    # 🧩 If the model outputs something too short, expand gracefully
+    if len(answer.split()) < 8:
+        answer = (
+            "The document mentions this briefly. Based on the context, here's what it suggests: "
+            + answer
         )
-        return result[0]["generated_text"].strip()
-    except Exception as e:
-        print(f"⚠️ Generation failed: {e}")
-        return "⚠️ Error: Could not generate an answer at the moment."
+
+    return answer
+
+except Exception as e:
+    print(f"⚠️ Generation failed: {e}")
+    return "⚠️ Error: Could not generate an answer at the moment."
+
 
 
 # ==========================================================